Classification of Extremist by Topic Modeling

Hafsa Bano; Dr Wasif Akbar; Dr Naeem Aslam; Muhammad Bilal

Author(s)

Hafsa Bano Department of Computer Science, NFC,IET of Engeerning and Technology,Multan, Punjab,Pakistan

Dr Wasif Akbar Department of Computer Science, NFC,IET of Engeerning and Technology,Multan, Punjab,Pakistan

Dr Naeem Aslam Department of Computer Science, NFC,IET of Engeerning and Technology,Multan, Punjab,Pakistan

Muhammad Bilal Department of Computer Science, NFC,IET of Engeerning and Technology,Multan, Punjab,Pakistan

Abstract

Social Media forums were formerly seen as the promised land for amusement, education, and boosting purposes that energize the evolution of text analysis to acknowledge the complications of real world. A seedbed for toxic conduct, fanatical content, and political propaganda, they have now become a sump. YouTube is one of the finest platforms where millions of comments thrown per day which consists of valuable and misleading information. Comments may contain slang, foreign and misspelled words that's very laborious to perform natural language processing (NLP) with those informal languages. Most preceding research diffused the analysis of gigantic unstructured data that comes in different formats and not simply classify in current databases. This study aims to analyze dynamic textual data from randomly selected channels on YouTube. For this ambition YouTube information API v3 was used to scrape a variety of data from YouTube videos. We investigated a few hot topics that are having mordant remarks about racism, LGBTQ, sub teen life aspect, and women and girls are just some of the areas where bullying is ordinary. In this study we summarize a custom dataset of 50 video's comments related to rap songs and other social events to perform topic modelling sentiment analysis that dwindle the cost of data labeling and annotating. The main objective of our research is to identify and classify each word with their probability from the dataset by applying model hyper parameters alpha, beta, gamma and examine the topic coherence score to measure the semantic similarity of words from extracted topics. This study's experimental results reveal that the analysis was able to achieve significant sentiment analysis efficiency at both the document and word levels.

Publication Details

Page(s) 235-248

DOI DOI not available

Published Journal: VFAST Transactions on Software Engineering, Volume: 11, Issue: 2, Year: 2023

Request the full text from PASTIC

Keywords

Text Mining, Topic Modelling, Unstructured Data, YouTube APIv3, Semantic Structure, Coherence Analysis